This paper describes a real-time implementation of a text-to-speech synthetizer, based on a diphone concatenation approach and running on a single board hardware which is plugged in a PC slot. The system is based on a 68000 microprocessor and two TMS32010; therefore the system can handle two channels simultaneously. The linguistic processing on the input ASCII string (i.e. text to diphone conversion and prosodic processing) is performed by the 68000 while the TMS32010 performs the actual speech synthesis by means of an LPC-12. The system is provided with a telephone interface, giving the possibility of remote listening of texts.